论文部分内容阅读
勘探开发初期海上油田钻井少、井间距离大,在应用地震多属性分析技术预测储层参数的过程中,直接采用监督最小二乘支持向量机算法预测精度较低。本文将最小二乘支持向量机与半监督学习理论结合,提出基于最小二乘支持向量机协同训练的半监督回归模型,并在模型训练过程中引入矩阵迭代求逆的方法,提高模型训练速度。利用UCI数据集实验研究,对比了半监督与监督最小二乘支持向量机模型,结果表明,半监督学习机制能够有效地提高最小二乘支持向量机的泛化性能,且随着训练样本的减小,效果更加明显;同时对比了半监督最小二乘支持向量机与半监督k-临近算法,结果显示,在小样本建模中,半监督最小二乘支持向量机有着更高的预测精度。最终将半监督最小二乘支持向量机运用于锦州工区,预测该区的砂体及储层孔隙度的分布,获得了较好的地质效果。
In the early stage of exploration and development, offshore oilfields lacked drilling and the distance between wells was large. In the process of predicting reservoir parameters by using seismic multi-attribute analysis technique, the prediction accuracy of the least square support vector machine directly using supervised LS-SVM is low. In this paper, the least square support vector machine and semi-supervised learning theory are combined to propose a semi-supervised regression model based on LS-SVM cooperative training, and the matrix iterative inversion method is introduced to improve the model training speed. The experimental results show that the semi-supervised learning mechanism can effectively improve the generalization performance of least square support vector machines, and as the training samples are reduced The results show that semi-supervised least square support vector machine has higher prediction accuracy in small sample modeling. Finally, the semi-supervised least square support vector machine is applied to Jinzhou work area to predict the distribution of sand body and reservoir porosity in this area and get better geological results.